Skip to content

Commit 3a248ed

Browse files
mhjensenmhjensen
committed
update
1 parent f60e0bf commit 3a248ed

8 files changed

Lines changed: 203 additions & 310 deletions

File tree

‎doc/pub/week13/html/week13-bs.html‎

Lines changed: 11 additions & 33 deletions
Original file line numberDiff line numberDiff line change
@@ -150,10 +150,6 @@
150150
2,
151151
None,
152152
'diffusion-models-for-hierarchical-vae-from-url-https-arxiv-org-abs-2208-11970'),
153-
('Equation for the Markovian hierarchical VAE',
154-
2,
155-
None,
156-
'equation-for-the-markovian-hierarchical-vae'),
157153
('Variational Diffusion Models',
158154
2,
159155
None,
@@ -166,10 +162,6 @@
166162
('Encoder transitions', 2, None, 'encoder-transitions'),
167163
('Third assumption', 2, None, 'third-assumption'),
168164
('Noisification', 2, None, 'noisification'),
169-
('Diffusion models, from URL:"https://arxiv.org/abs/2208.11970"',
170-
2,
171-
None,
172-
'diffusion-models-from-url-https-arxiv-org-abs-2208-11970'),
173165
('Gaussian modeling', 2, None, 'gaussian-modeling'),
174166
('Optimizing the variational diffusion model',
175167
2,
@@ -270,14 +262,12 @@
270262
<!-- navigation toc: --> <li><a href="#mathematical-representation" style="font-size: 80%;">Mathematical representation</a></li>
271263
<!-- navigation toc: --> <li><a href="#back-to-the-marginal-probability" style="font-size: 80%;">Back to the marginal probability</a></li>
272264
<!-- navigation toc: --> <li><a href="#diffusion-models-for-hierarchical-vae-from-url-https-arxiv-org-abs-2208-11970" style="font-size: 80%;">Diffusion models for hierarchical VAE, from URL:"https://arxiv.org/abs/2208.11970"</a></li>
273-
<!-- navigation toc: --> <li><a href="#equation-for-the-markovian-hierarchical-vae" style="font-size: 80%;">Equation for the Markovian hierarchical VAE</a></li>
274265
<!-- navigation toc: --> <li><a href="#variational-diffusion-models" style="font-size: 80%;">Variational Diffusion Models</a></li>
275266
<!-- navigation toc: --> <li><a href="#second-assumption" style="font-size: 80%;">Second assumption</a></li>
276267
<!-- navigation toc: --> <li><a href="#parameterizing-gaussian-encoder" style="font-size: 80%;">Parameterizing Gaussian encoder</a></li>
277268
<!-- navigation toc: --> <li><a href="#encoder-transitions" style="font-size: 80%;">Encoder transitions</a></li>
278269
<!-- navigation toc: --> <li><a href="#third-assumption" style="font-size: 80%;">Third assumption</a></li>
279270
<!-- navigation toc: --> <li><a href="#noisification" style="font-size: 80%;">Noisification</a></li>
280-
<!-- navigation toc: --> <li><a href="#diffusion-models-from-url-https-arxiv-org-abs-2208-11970" style="font-size: 80%;">Diffusion models, from URL:"https://arxiv.org/abs/2208.11970"</a></li>
281271
<!-- navigation toc: --> <li><a href="#gaussian-modeling" style="font-size: 80%;">Gaussian modeling</a></li>
282272
<!-- navigation toc: --> <li><a href="#optimizing-the-variational-diffusion-model" style="font-size: 80%;">Optimizing the variational diffusion model</a></li>
283273
<!-- navigation toc: --> <li><a href="#continues" style="font-size: 80%;">Continues</a></li>
@@ -1373,9 +1363,10 @@ <h2 id="diffusion-learning" class="anchor">Diffusion learning </h2>
13731363
<h2 id="mathematics-of-diffusion-models" class="anchor">Mathematics of diffusion models </h2>
13741364

13751365
<p>Let us go back our discussions of the variational autoencoders from
1376-
last week, see
1377-
<a href="https://github.com/CompPhysics/AdvancedMachineLearning/blob/main/doc/pub/week12/ipynb/week12.ipynb" target="_self"><tt>https://github.com/CompPhysics/AdvancedMachineLearning/blob/main/doc/pub/week12/ipynb/week12.ipynb</tt></a>. As
1378-
a first attempt at understanding diffusion models, we can think of
1366+
last week and above.
1367+
</p>
1368+
1369+
<p>As a first attempt at understanding diffusion models, we can think of
13791370
these as stacked VAEs, or better, recursive VAEs.
13801371
</p>
13811372

@@ -1438,15 +1429,6 @@ <h2 id="diffusion-models-for-hierarchical-vae-from-url-https-arxiv-org-abs-2208-
14381429
\( \boldsymbol{h}_{t+1} \). Here \( \boldsymbol{z} \) is our latent variable \( \boldsymbol{h} \).
14391430
</p>
14401431

1441-
<br/><br/>
1442-
<center>
1443-
<p><img src="figures/figure1.png" width="800" align="bottom"></p>
1444-
</center>
1445-
<br/><br/>
1446-
1447-
<!-- !split -->
1448-
<h2 id="equation-for-the-markovian-hierarchical-vae" class="anchor">Equation for the Markovian hierarchical VAE </h2>
1449-
14501432
<p>We obtain then</p>
14511433
$$
14521434
\begin{align*}
@@ -1553,15 +1535,6 @@ <h2 id="noisification" class="anchor">Noisification </h2>
15531535
identical to pure Gaussian noise. See figure on next slide.
15541536
</p>
15551537

1556-
<!-- !split -->
1557-
<h2 id="diffusion-models-from-url-https-arxiv-org-abs-2208-11970" class="anchor">Diffusion models, from <a href="https://arxiv.org/abs/2208.11970" target="_self"><tt>https://arxiv.org/abs/2208.11970</tt></a> </h2>
1558-
1559-
<br/><br/>
1560-
<center>
1561-
<p><img src="figures/figure2.png" width="800" align="bottom"></p>
1562-
</center>
1563-
<br/><br/>
1564-
15651538
<!-- !split -->
15661539
<h2 id="gaussian-modeling" class="anchor">Gaussian modeling </h2>
15671540

@@ -1655,8 +1628,13 @@ <h2 id="more-details" class="anchor">More details </h2>
16551628

16561629
<p>For more details and implementaions, see Calvin Luo at <a href="https://arxiv.org/abs/2208.11970" target="_self"><tt>https://arxiv.org/abs/2208.11970</tt></a></p>
16571630

1658-
<!-- FIGURE: [figures/figure4.png, width=800 frac=1.0] -->
1659-
1631+
<p>The plans for the rest of the semester are</p>
1632+
<ol>
1633+
<li> May 1: public holiday</li>
1634+
<li> May 8: More on diffusion models and introduction to GANs. Project work</li>
1635+
<li> May 15: GANS and summary of course. Project work. This could also be the last lecture</li>
1636+
<li> May 22: Alternatively, we could discuss briefly reinforcement learning and do a summary of the course. Project work.</li>
1637+
</ol>
16601638
<!-- ------------------- end of main content --------------- -->
16611639
</div> <!-- end container -->
16621640
<!-- include javascript, jQuery *first* -->

‎doc/pub/week13/html/week13-reveal.html‎

Lines changed: 11 additions & 24 deletions
Original file line numberDiff line numberDiff line change
@@ -1352,9 +1352,10 @@ <h2 id="diffusion-learning">Diffusion learning </h2>
13521352
<h2 id="mathematics-of-diffusion-models">Mathematics of diffusion models </h2>
13531353

13541354
<p>Let us go back our discussions of the variational autoencoders from
1355-
last week, see
1356-
<a href="https://github.com/CompPhysics/AdvancedMachineLearning/blob/main/doc/pub/week12/ipynb/week12.ipynb" target="_blank"><tt>https://github.com/CompPhysics/AdvancedMachineLearning/blob/main/doc/pub/week12/ipynb/week12.ipynb</tt></a>. As
1357-
a first attempt at understanding diffusion models, we can think of
1355+
last week and above.
1356+
</p>
1357+
1358+
<p>As a first attempt at understanding diffusion models, we can think of
13581359
these as stacked VAEs, or better, recursive VAEs.
13591360
</p>
13601361

@@ -1421,16 +1422,6 @@ <h2 id="diffusion-models-for-hierarchical-vae-from-url-https-arxiv-org-abs-2208-
14211422
\( \boldsymbol{h}_{t+1} \). Here \( \boldsymbol{z} \) is our latent variable \( \boldsymbol{h} \).
14221423
</p>
14231424

1424-
<br/><br/>
1425-
<center>
1426-
<p><img src="figures/figure1.png" width="800" align="bottom"></p>
1427-
</center>
1428-
<br/><br/>
1429-
</section>
1430-
1431-
<section>
1432-
<h2 id="equation-for-the-markovian-hierarchical-vae">Equation for the Markovian hierarchical VAE </h2>
1433-
14341425
<p>We obtain then</p>
14351426
<p>&nbsp;<br>
14361427
$$
@@ -1547,16 +1538,6 @@ <h2 id="noisification">Noisification </h2>
15471538
</p>
15481539
</section>
15491540

1550-
<section>
1551-
<h2 id="diffusion-models-from-url-https-arxiv-org-abs-2208-11970">Diffusion models, from <a href="https://arxiv.org/abs/2208.11970" target="_blank"><tt>https://arxiv.org/abs/2208.11970</tt></a> </h2>
1552-
1553-
<br/><br/>
1554-
<center>
1555-
<p><img src="figures/figure2.png" width="800" align="bottom"></p>
1556-
</center>
1557-
<br/><br/>
1558-
</section>
1559-
15601541
<section>
15611542
<h2 id="gaussian-modeling">Gaussian modeling </h2>
15621543

@@ -1661,7 +1642,13 @@ <h2 id="more-details">More details </h2>
16611642

16621643
<p>For more details and implementaions, see Calvin Luo at <a href="https://arxiv.org/abs/2208.11970" target="_blank"><tt>https://arxiv.org/abs/2208.11970</tt></a></p>
16631644

1664-
<!-- FIGURE: [figures/figure4.png, width=800 frac=1.0] -->
1645+
<p>The plans for the rest of the semester are</p>
1646+
<ol>
1647+
<p><li> May 1: public holiday</li>
1648+
<p><li> May 8: More on diffusion models and introduction to GANs. Project work</li>
1649+
<p><li> May 15: GANS and summary of course. Project work. This could also be the last lecture</li>
1650+
<p><li> May 22: Alternatively, we could discuss briefly reinforcement learning and do a summary of the course. Project work.</li>
1651+
</ol>
16651652
</section>
16661653

16671654

‎doc/pub/week13/html/week13-solarized.html‎

Lines changed: 11 additions & 31 deletions
Original file line numberDiff line numberDiff line change
@@ -177,10 +177,6 @@
177177
2,
178178
None,
179179
'diffusion-models-for-hierarchical-vae-from-url-https-arxiv-org-abs-2208-11970'),
180-
('Equation for the Markovian hierarchical VAE',
181-
2,
182-
None,
183-
'equation-for-the-markovian-hierarchical-vae'),
184180
('Variational Diffusion Models',
185181
2,
186182
None,
@@ -193,10 +189,6 @@
193189
('Encoder transitions', 2, None, 'encoder-transitions'),
194190
('Third assumption', 2, None, 'third-assumption'),
195191
('Noisification', 2, None, 'noisification'),
196-
('Diffusion models, from URL:"https://arxiv.org/abs/2208.11970"',
197-
2,
198-
None,
199-
'diffusion-models-from-url-https-arxiv-org-abs-2208-11970'),
200192
('Gaussian modeling', 2, None, 'gaussian-modeling'),
201193
('Optimizing the variational diffusion model',
202194
2,
@@ -1304,9 +1296,10 @@ <h2 id="diffusion-learning">Diffusion learning </h2>
13041296
<h2 id="mathematics-of-diffusion-models">Mathematics of diffusion models </h2>
13051297

13061298
<p>Let us go back our discussions of the variational autoencoders from
1307-
last week, see
1308-
<a href="https://github.com/CompPhysics/AdvancedMachineLearning/blob/main/doc/pub/week12/ipynb/week12.ipynb" target="_blank"><tt>https://github.com/CompPhysics/AdvancedMachineLearning/blob/main/doc/pub/week12/ipynb/week12.ipynb</tt></a>. As
1309-
a first attempt at understanding diffusion models, we can think of
1299+
last week and above.
1300+
</p>
1301+
1302+
<p>As a first attempt at understanding diffusion models, we can think of
13101303
these as stacked VAEs, or better, recursive VAEs.
13111304
</p>
13121305

@@ -1369,15 +1362,6 @@ <h2 id="diffusion-models-for-hierarchical-vae-from-url-https-arxiv-org-abs-2208-
13691362
\( \boldsymbol{h}_{t+1} \). Here \( \boldsymbol{z} \) is our latent variable \( \boldsymbol{h} \).
13701363
</p>
13711364

1372-
<br/><br/>
1373-
<center>
1374-
<p><img src="figures/figure1.png" width="800" align="bottom"></p>
1375-
</center>
1376-
<br/><br/>
1377-
1378-
<!-- !split --><br><br><br><br><br><br><br><br><br><br>
1379-
<h2 id="equation-for-the-markovian-hierarchical-vae">Equation for the Markovian hierarchical VAE </h2>
1380-
13811365
<p>We obtain then</p>
13821366
$$
13831367
\begin{align*}
@@ -1482,15 +1466,6 @@ <h2 id="noisification">Noisification </h2>
14821466
identical to pure Gaussian noise. See figure on next slide.
14831467
</p>
14841468

1485-
<!-- !split --><br><br><br><br><br><br><br><br><br><br>
1486-
<h2 id="diffusion-models-from-url-https-arxiv-org-abs-2208-11970">Diffusion models, from <a href="https://arxiv.org/abs/2208.11970" target="_blank"><tt>https://arxiv.org/abs/2208.11970</tt></a> </h2>
1487-
1488-
<br/><br/>
1489-
<center>
1490-
<p><img src="figures/figure2.png" width="800" align="bottom"></p>
1491-
</center>
1492-
<br/><br/>
1493-
14941469
<!-- !split --><br><br><br><br><br><br><br><br><br><br>
14951470
<h2 id="gaussian-modeling">Gaussian modeling </h2>
14961471

@@ -1584,8 +1559,13 @@ <h2 id="more-details">More details </h2>
15841559

15851560
<p>For more details and implementaions, see Calvin Luo at <a href="https://arxiv.org/abs/2208.11970" target="_blank"><tt>https://arxiv.org/abs/2208.11970</tt></a></p>
15861561

1587-
<!-- FIGURE: [figures/figure4.png, width=800 frac=1.0] -->
1588-
1562+
<p>The plans for the rest of the semester are</p>
1563+
<ol>
1564+
<li> May 1: public holiday</li>
1565+
<li> May 8: More on diffusion models and introduction to GANs. Project work</li>
1566+
<li> May 15: GANS and summary of course. Project work. This could also be the last lecture</li>
1567+
<li> May 22: Alternatively, we could discuss briefly reinforcement learning and do a summary of the course. Project work.</li>
1568+
</ol>
15891569
<!-- ------------------- end of main content --------------- -->
15901570
<center style="font-size:80%">
15911571
<!-- copyright --> &copy; 1999-2025, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license

‎doc/pub/week13/html/week13.html‎

Lines changed: 11 additions & 31 deletions
Original file line numberDiff line numberDiff line change
@@ -254,10 +254,6 @@
254254
2,
255255
None,
256256
'diffusion-models-for-hierarchical-vae-from-url-https-arxiv-org-abs-2208-11970'),
257-
('Equation for the Markovian hierarchical VAE',
258-
2,
259-
None,
260-
'equation-for-the-markovian-hierarchical-vae'),
261257
('Variational Diffusion Models',
262258
2,
263259
None,
@@ -270,10 +266,6 @@
270266
('Encoder transitions', 2, None, 'encoder-transitions'),
271267
('Third assumption', 2, None, 'third-assumption'),
272268
('Noisification', 2, None, 'noisification'),
273-
('Diffusion models, from URL:"https://arxiv.org/abs/2208.11970"',
274-
2,
275-
None,
276-
'diffusion-models-from-url-https-arxiv-org-abs-2208-11970'),
277269
('Gaussian modeling', 2, None, 'gaussian-modeling'),
278270
('Optimizing the variational diffusion model',
279271
2,
@@ -1381,9 +1373,10 @@ <h2 id="diffusion-learning">Diffusion learning </h2>
13811373
<h2 id="mathematics-of-diffusion-models">Mathematics of diffusion models </h2>
13821374

13831375
<p>Let us go back our discussions of the variational autoencoders from
1384-
last week, see
1385-
<a href="https://github.com/CompPhysics/AdvancedMachineLearning/blob/main/doc/pub/week12/ipynb/week12.ipynb" target="_blank"><tt>https://github.com/CompPhysics/AdvancedMachineLearning/blob/main/doc/pub/week12/ipynb/week12.ipynb</tt></a>. As
1386-
a first attempt at understanding diffusion models, we can think of
1376+
last week and above.
1377+
</p>
1378+
1379+
<p>As a first attempt at understanding diffusion models, we can think of
13871380
these as stacked VAEs, or better, recursive VAEs.
13881381
</p>
13891382

@@ -1446,15 +1439,6 @@ <h2 id="diffusion-models-for-hierarchical-vae-from-url-https-arxiv-org-abs-2208-
14461439
\( \boldsymbol{h}_{t+1} \). Here \( \boldsymbol{z} \) is our latent variable \( \boldsymbol{h} \).
14471440
</p>
14481441

1449-
<br/><br/>
1450-
<center>
1451-
<p><img src="figures/figure1.png" width="800" align="bottom"></p>
1452-
</center>
1453-
<br/><br/>
1454-
1455-
<!-- !split --><br><br><br><br><br><br><br><br><br><br>
1456-
<h2 id="equation-for-the-markovian-hierarchical-vae">Equation for the Markovian hierarchical VAE </h2>
1457-
14581442
<p>We obtain then</p>
14591443
$$
14601444
\begin{align*}
@@ -1559,15 +1543,6 @@ <h2 id="noisification">Noisification </h2>
15591543
identical to pure Gaussian noise. See figure on next slide.
15601544
</p>
15611545

1562-
<!-- !split --><br><br><br><br><br><br><br><br><br><br>
1563-
<h2 id="diffusion-models-from-url-https-arxiv-org-abs-2208-11970">Diffusion models, from <a href="https://arxiv.org/abs/2208.11970" target="_blank"><tt>https://arxiv.org/abs/2208.11970</tt></a> </h2>
1564-
1565-
<br/><br/>
1566-
<center>
1567-
<p><img src="figures/figure2.png" width="800" align="bottom"></p>
1568-
</center>
1569-
<br/><br/>
1570-
15711546
<!-- !split --><br><br><br><br><br><br><br><br><br><br>
15721547
<h2 id="gaussian-modeling">Gaussian modeling </h2>
15731548

@@ -1661,8 +1636,13 @@ <h2 id="more-details">More details </h2>
16611636

16621637
<p>For more details and implementaions, see Calvin Luo at <a href="https://arxiv.org/abs/2208.11970" target="_blank"><tt>https://arxiv.org/abs/2208.11970</tt></a></p>
16631638

1664-
<!-- FIGURE: [figures/figure4.png, width=800 frac=1.0] -->
1665-
1639+
<p>The plans for the rest of the semester are</p>
1640+
<ol>
1641+
<li> May 1: public holiday</li>
1642+
<li> May 8: More on diffusion models and introduction to GANs. Project work</li>
1643+
<li> May 15: GANS and summary of course. Project work. This could also be the last lecture</li>
1644+
<li> May 22: Alternatively, we could discuss briefly reinforcement learning and do a summary of the course. Project work.</li>
1645+
</ol>
16661646
<!-- ------------------- end of main content --------------- -->
16671647
<center style="font-size:80%">
16681648
<!-- copyright --> &copy; 1999-2025, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license

‎doc/pub/week13/ipynb/ipynb-week13-src.tar.gz‎

-3.68 MB
Binary file not shown.

0 commit comments

Comments
 (0)